Sorry, I have listened to most top CAS servers in the market, including the Wadax Atlantis.
To me, they can be divided into 2 categories/leagues :
1. The Extreme.
2. The non-Extremes.
It's that simple!
You are welcome but it was a joke for Taiko...
Taiko Audio SGM Extreme : the Crème de la Crème
- Thread starter CKKeung
- Start date
You are welcome but it was a joke for Taiko...
You forgot Wadax Atlantis Server.
Even Robert Koda kneels in front of such magnificent electronic....
View attachment 59798
Actually I think Robert is stunned at the beautiful industrial design
His 160's look awesome...
Thanks for the reply! Could use this machine also for more than just music with that amount of beastly RAM
Actually I think Robert is stunned at the beautiful industrial design
His 160's look awesome...
Hi Taiko.
Why is it necessary to have an excess computing power if what I am looking for is to deliver an unaltered, unprocessed data?
I don't need a large computing power to play audio unless I want to process a large amount of data.
This last case would be necessary to play from HQ Player.
But what advantages HQ player has i if i am altering the audio signal?
Why i need to introduce processes that are only important in the Dac processor?
The computing power must be within the Dac based on a result, and that result must be given by the analog digital conversion quality where the Dac processor has the real and necessary utility of audio signal processing, not on the server processor.
Regards.
Last edited:
To reduce the RF noise signature of the music playback, going multicore is an effective strategy. The execution time is reduced, and processes need to wait much less often for a previous process to complete.
Emile has tested and I have heard Extremes fitted with CPU’s with a lower core count, and they sounded noticeably inferior to the 10 core CPU’s which are now used
Thank you for your answer.
Sincerely, i don’t believe this kind of trial and error design.
Since i bought my server on 2017 i’ve seen three SGM servers.
SGM 2015 - Intel Skylake i7 6700K
SGM Evo - Intel Skylake i7 7700K
SGM Extreme - Dual Intel Xeon Scalable 10 core
Could we expect a new SGM edition with new 2020 Intel Comet Lake-S next april?
Based on you logic, it will sound better.
That isn’t an effective strategy on short time.
You says going multicore is better for RF noise? And worst for nothing?
I don’t see any grounding post to get better EMI/RF noise reduction for the most noisy component. The power supply.
Execution time is reduced? What kind of execution? Bit Perfect audio process execution is very simple for any processor.
As i said, the processor is only relevant on the Dac.
I’m sorry to disagree.
Cheers.
Sincerely, i don’t believe this kind of trial and error design.
Since i bought my server on 2017 i’ve seen three SGM servers.
SGM 2015 - Intel Skylake i7 6700K
SGM Evo - Intel Skylake i7 7700K
SGM Extreme - Dual Intel Xeon Scalable 10 core
Could we expect a new SGM edition with new 2020 Intel Comet Lake-S next april?
Based on you logic, it will sound better.
That isn’t an effective strategy on short time.
You says going multicore is better for RF noise? And worst for nothing?
I don’t see any grounding post to get better EMI/RF noise reduction for the most noisy component. The power supply.
Execution time is reduced? What kind of execution? Bit Perfect audio process execution is very simple for any processor.
As i said, the processor is only relevant on the Dac.
I’m sorry to disagree.
Cheers.
Last edited:
Surely digital audio is always going to be processor dependent? I don’t think anyone is under the illusion that digital stands still. That is any analog is so much more comforting. IMO, an upgrade path is important and Emile’s line up seems to provide that to some extent.
You are of course fully entitled to your opinion.
The SGM2015 and its EVO update were designed with HQ Player upsampling in mind. HQ Player favours core clock speeds over core count. These use consumer level CPUs which are replaced quite often, and are unfortunately neither drop in replacements, nor stay available.
The SGM Extreme was designed with “bit perfect” playback in mind. It uses professional Xeon class parts, if new generations are not drop in replacements the previous will be kept available for 10 years.
We did not get increased benefits from going past 2*10 cores but have tested up to 2*20.
Roon Core is a very busy piece of software with a lot of threads.
A very audible part of the Extreme SQ Is the memory architecture. This is a feature we can expect to stay the best strategy for quite a while
If you can suggest a better way for development than laborious Trial and Error, we are listening
You asked, so here’s my opinion
Trial and error is probably the worst way to develop a piece of technology. It basically infers that you keep trying stuff until you find something that works, or not. When do you stop? At what point is your piece of hardware ready? Finished? If you are about to release and a new very promising piece of kit releases, do you hold your release pending another trial? Or do you release knowing that it could, maybe be better, or not?
What is the design goal? If its best sound, then why doesn’t the Extreme include for example SoTA footers, fuses, power cables? The answer I’m guessing is ‘because those things sound different in every system, so we leave it up to our customers to listen and decide what they need‘.
The above sounds like a condemnation of the Extreme...its not. The Extreme is a fabulous piece of kit. You asked about trial and error development. What trial and error should do, if you’re thoroughly committed to the process is give you a product that ultimately kicks ass. Essentially you’ve tried everything and you've picked the best. So ultimately the process should deliver. However there‘s a caveat. When you’ve developed something in this fashion, you can’t really reach any generalized conclusions about it like “10 core processors are better“ . The conclusion you can reach is that in the context of the Extreme‘s architecture and design choices, 10 core CPUs delivered the best sound quality. The reason for the qualification is manifold;
A different front end strategy i.e bringing in a cleaner stream impacts the sound and certainly influences CPU load. Roon and all its background processing influences the sound and the CPU load. Windows 10 influences sound and CPU load, so its not surprising that the Extreme benefits from plenty of CPU horsepower. Another design that’s far more economical with CPU resources could be quite different.
If i were to follow the usual high tech development principles, the approach would be different. First a design map would be developed, outlining all the desired design goals with regard to everything that you know and research tells you influences sound quality, namely:
Noise....internal noise generation and external radiated and conducted noise. The design should seek to minimize all sources and make the system as immune as possible. Checking noise levels would be a benchmark measure during development. Every design decision you make would be predicated on its noise profile.
Vibration.....internally and external generated sources. Simlar to the above, the design should seek to minimize vibration of internal components, finding ways to minimize production of vibration and ground or neutralise what can’t be avoided. Ways of minimising the effects of external vibration are part of the design and again, vibration would be a key bench measurement during development
Power.....here the design would seek to minimize the interplay of various components through the power supplies, would seek to avoid the generation of noise and would assure the greatest degree of isolation and stability for key components. In addition is would seek to make the unit immune to incoming power noise and would assure the unit itself wasn’t generating problems for other components by injecting its own noise into ground planes etc. Power quality would also be a benchmark measure during development.
Software......This is of course a major area for concern.....use commercially available software, with its typically massive overkill in terms of unused and therefore unwanted features, or sign up to the difficult, time consuming and costly own development. On the one hand, you save time and money but get something that is by definition only a moderately good fit to your specific needs, while on the other hand you get something that exactly fits your needs, assuming you can actually pull off the development..,.,..which is by no means a trivial exercise.
Obviously there’s more but you get my drift.
Finally, with trial and error design, you have to throw out any concept of ‘target price’. I’m not saying this is a good or bad thing, just that you have no control over price whatsoever, so while you‘re building, you’ve no real idea how much your product is going to sell for. I do realise that this seems to be a less important element of audiophile product development, but the risk is always that you end up with a product that costs a lot more but doesn’t sound a lot better than competitors. I’m talking risk here, not reality. (I do get it that the Extreme SQ is SoTA
)Anyway, I’m sure you get the gist of my message. Trial and error is a good way to finish, fine tune or ‘polish’ something, but doesn’t really suit ground-up development, simply because it lacks specific direction and specified outcomes. With trial and error, you’re never quite sure when you’ve actually arrived. Its a bit like going for a hike in the mountains and not taking a map.......just following what looks like the best trails and knowing the general direction you should be headed.
Last edited:
You do make some good points, but not every direction is known when it comes to digital. For example, it took trial and error to realize just how incredibly sensitive the human ear is to digital jitter (as opposed to analog wow and flutter). Nobody knew in advance when digital was developed.
So it is good to have a specific direction, but it is equally important to be open to the as of yet unknown.
I also doubt that Emile did such a crude trial and error methodology as has been suggested. Whilst it would be my method given a complete lack of knowledge in this field, i'm sure the Extreme basics was probably constructed based on predicated known parameters and systems a priori but the final result required extensive testing of those areas that were indeed unknown. The extreme does not strike me as the kind of product you get just by simply throwing shit at the wall and seeing what sticks.
That’s of course the opposite side of the argument and IMO no less valid. There’s a place for basic research to discover where the pressure points lie. No doubt at all about that.
I’m not an Extreme user and I haven’t even heard one, so I’m not in the least qualified to speak about its sound quality, so nothing I say relates to its actual performance, although ears I very much trust tell me its really great (Just to be clear).
I have an Innuos Statement and I bought that without hearing it first. Why? Because I completely trusted the engineering. I had a Zenith SE, which sounded great from the get go...pretty much better than anything digital I’d heard to that point and it met all my needs including ripping a fairly substantial CD collection. The Zenith SE was built with sound quality and certain features in mind which met my needs, hence the purchase. When i looked at the engineering, power supplies, vibration control, EMI and RFI minimisation, processor efficiency and software for playing and managing music were at the top of Innuos’s development list....things I already knew to make a potentially massive difference when implemented optimally. I wasn’t disappointed so when the Statement was announced I saw that all the ‘pressure points‘ had been further refined and optimised, to quite a degree. Based on that judgement alone I ordered a unit. Given the results I was getting with the Innuos products, I wondered what applying the exact same engineering pressure points to the network feed would do. Power supplies, EMI and RFI screening, Isolation against conducted EMI, Vibration control etc. They were all there in the network feed, so I did just that. My network now runs exclusively with a particular cable brand known for its screening prowess, all powered network points have their own dedicated high quality power supplies and highly screened DC cables of optimum length, all networking modules and power supplies are on vibration minimising mounts or platforms, the mains power supply is optimised with dedicated radials and power cables and I have avoided any EMI generating sources in the hi-fi room, where possible. The result? Stunning and worth every penny. Every single step along the way rewarded in substantial SQ improvements, so it was easy maintaining faith and making further investments
My conclusion? The Statement is a great piece of kit, ditto the Extreme but neither are going to produce their absolute best unless they are placed in an environment that reflects and extends or continues their design goals. I would suggest that using trial and error to create that environment would be expensive, time consuming and possibly not result in the best sound. Creating the environment based on sound engineering decisions will IMO get you there quicker and more certainly.
Now for where your argument comes in: Fibre Optics! If I never try it, I’ll never know. It may be better. It could be worse, but unless I give it a try, who knows? But one thing I do know.....this journey of discovery is a lot of fun!!
Hey Howiebrou, if throwing shit at the wall was how my post came across, then I certainly wasn’t sufficiently clear. You are absolutely correct in that a product like the Extreme doesn’t come from a careless development cycle. I’m sure there were R&D breadboards and prototypes where things like memory, CPUs etc could be thoroughly evaluated and compared.
My posts was really about Trial & Error product development and really not about the Extreme. I was pointing out that Trial and Error is not a good way to proceed for a number of reasons already given in the previous thread. I was not privy to the Extreme’s development and I don’t know how it sounds (more’s the pity) so my comments were about a design philosophy rather than about a product. Is that a bit clearer?
Late edit.....with the greatest respect, I’m thinking that Trial & Error was maybe not a good description given its negative connotation. For the Extreme it seems like it was more a multiple trial development, where each stage was thoroughly evaluated to find the best for that particular application. That’s not really trial and error, which is typically a much less rigid set of trials. One looks for something that does the job, the other seeks to find the best component for the job. Big difference
Last edited:
